10.07.2015 Views

Web Mining and Social Networking: Techniques and ... - tud.ttu.ee

Web Mining and Social Networking: Techniques and ... - tud.ttu.ee

Web Mining and Social Networking: Techniques and ... - tud.ttu.ee

SHOW MORE
SHOW LESS
  • No tags were found...

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

52 3 Algorithms <strong>and</strong> <strong>Techniques</strong>uses these to modify its weights, <strong>and</strong> it aims to match its classifications with the targets in thetraining set. Weight modification follows a learning rule. The pseudo code of training neuralnetwork is shown in Algorithm 3.6.Algorithm 3.6: Training neural network algorithmfully trained = FALSE;While fully trained != TRUE dofully trained = TRUE;For each training vector = < X 1 ,X 2 ,...,X n ,θ,target > doa = (X 1 ∗W 1 )+(X 2 ∗W 2 )+...+(X n ∗W n ) − θ;y = sigmoid(a);If y != target then fully trained = FALSE;For each W i doMODIFY WEIGHT( W i );endendIf fully trained then BREAK;endThe advantage of Neural nets is that it performs very well on difficult, non-linear domains,where it becomes more <strong>and</strong> more difficult to use Decision tr<strong>ee</strong>, or Rule induction systems,which cut the space of examples parallel to attribute axes. One of disadvantages in usingNeural nets for data mining is a slow learning process, compared top for example Decisiontr<strong>ee</strong>s. This difference can very easily be several orders of magni<strong>tud</strong>e. Another disadvantageis that neural networks do not give explicit knowledge representation in the form of rules, orsome other easily interpretable form. The model is implicit, hidden in the network structure<strong>and</strong> optimized weights, betw<strong>ee</strong>n the nodes.In addition to the above commonly used supervised learning methods, there are manyothers developed, such as Discriminative Classi f ication, Maximum Entropy Learners, SV M,<strong>and</strong> so forth. Refer [33] for a more comprehensive survey.3.3 Unsupervised LearningIn this section, we will introduce major techniques of unsupervised learning (or clustering).Among a large amount of approaches that have b<strong>ee</strong>n proposed, there are thr<strong>ee</strong> representativeunsupervised learning strategies, i.e., k-means, hierarchical clustering <strong>and</strong> density basedclustering.3.3.1 The k-Means AlgorithmK-means clustering algorithm was first proposed in [173, 110]. In briefly, k-means clusteringis a top-down algorithm that classifies the objects into k number of groups with regard toattributes or features, where k is a positive integer number <strong>and</strong> specified apriori by users.The grouping is done by minimizing the sum of squares of distances betw<strong>ee</strong>n object <strong>and</strong> thecorresponding cluster centroid.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!